Cell, Vol. 84, 5–8, January 12, 1996, Copyright 1996 by Cell Press
DNA Polymerase III:
Running Rings around the Fork
Daniel R. Herendeen and Thomas J. Kelly
Department of Molecular Biology and Genetics
The Johns Hopkins University School of Medicine
Baltimore, Maryland 21205
Metabolic processes are often orchestrated by the coordinated action of multiple protein components. Because
of the complexity of such enzymatic mechanisms, the
participant proteins are aptly referred to as constituting
enzymatic “machinery.” Deciphering the inner workings
of the multiprotein machines that mediate processes,
such as DNA replication and transcription, is a major
goal of biology, but is a technically demanding task
owing to the difficulty in reassembling functional complexes from purified components outside of the cell.
Since its discovery nearly 25 years ago, the replicase
of Escherichia coli, DNA polymerase III (pol III) holoenzyme, has been extensively studied as a model replication machine (Kornberg and Baker, 1992; Kelman and
O’Donnell, 1995). The 10 protein subunits of pol III holoenzyme function in cooperation with other replication
proteins to carry out the duplication of the entire 4.4 Mb
E. coli chromosome in 30–40 min. Over the past decade,
work in several laboratories resulted in the identification
of the genes encoding all 10 subunits and the high level
expression of the corresponding gene products. This
accomplishment has made possible elegant biochemical studies that have brought understanding of the structure and function of the pol III holoenzyme to a level of
detail unmatched by other protein machines.
At the E. coli replication fork, the DNA duplex is progressively unwound by the action of a DNA helicase,
and the exposed single strands serve as templates for
the synthesis of short RNA primers by the primase and
associated proteins. The role of pol III holoenzyme is to
elongate newly synthesized primers to generate the two
progeny strands. Because of the antiparallel nature of
the DNA duplex, two different modes of priming are
required. Polymerization of one progeny strand (the
“leading” strand) occurs in the same direction as the
replication fork moves. Thus, only a single priming event
is required, after which the leading strand is elongated
continuously by pol III holoenzyme. Leading strand synthesis is highly processive owing to the presence of a
“sliding clamp” subunit that tethers the polymerase to
the template. Polymerization of the second progeny
strand (the “lagging” strand) occurs in the direction opposite to replication fork movement. Thus, elongation
of the lagging strand is a discontinuous process involving the repeated synthesis of RNA primers that are then
extended into short DNA chains (Okazaki fragments) by
pol III holoenzyme. Completion of the lagging strand
requires a repair system to remove the primers, fill in
the resulting gaps, and join together the short nascent
DNA strands. It is likely that the synthesis of both the
leading and the lagging strands at a chromosomal replication fork is mediated by a single pol III holoenzyme
Minireview
molecule that contains two identical DNA polymerase
subunits (Johanson and McHenry, 1984) (see below).
The synthesis of the lagging strand by pol III holoenzyme is a complex process that entails a number of
discrete steps that must occur in an orderly and efficient
fashion. To complete the synthesis of the chromosome
within 30–40 min, RNA primers are generated on the
lagging strand template every 1–2 s at average intervals
of 1–2 kb. The elongation of each primer by pol III holoenzyme takes place at a rate of about 1000 nucleotides
per second and is highly processive owing to the presence of the sliding clamp subunit. The discontinuous
mode of replication demands that pol III must cycle to
the next RNA primer upon completion of each Okazaki
fragment. This raises two potential difficulties. First, the
cycling process must be very rapid, occupying only a
fraction of the total time devoted to polymerization.
Rapid cycling is essential to ensure that the synthesis
of the lagging strand keeps pace with the synthesis of
the leading strand. Second, the requirement for cycling
of pol III would appear, at least at first sight, to be at
odds with the highly processive character of the polymerization process. Recent experiments by O’Donnell
and colleagues suggest that these problems are solved
by a remarkable mechanism that involves the partial
disassembly and reassembly of the holoenzyme structure during the synthesis of each Okazaki fragment (Stukenberg et al., 1994; Naktinis et al., 1996 [this issue of
Cell]). The mechanism is powered by ATP hydrolysis
and is controlled by specific protein–protein and protein–DNA interactions.
Pol III holoenzyme is composed of 10 unique subunits
and harbors at least three essential enzymatic activities
(Table 1). The enzyme contains four distinct functional
components: the core polymerase (aeu), which contains
both DNA polymerase (a) and proofreading exonuclease
(e) activities; the sliding clamp (b dimer), which confers
processivity by tethering the holoenzyme to the template DNA; the clamp loader or g complex (g2d1d9 1x1c 1),
which assembles b clamps onto the DNA in an ATPdependent reaction; the linker protein (t2), which binds
two core polymerase molecules and one g complex. The
structure of a stable subassembly of pol III, known as
pol III*, has been studied in detail by a variety of methods. Pol III* contains two core polymerases, one t dimer
and one g complex (Figure 1). The enzyme exhibits
greatly reduced processivity relative to the holoenzyme
because it lacks the b subunit, which readily dissociates
from the holoenzyme during purification. Addition of the
b subunit to pol III* regenerates the holoenzyme and
restores processivity. The pol III* complex can be reconstituted from individual subunits, and a general picture
of its overall organization has been deduced from detailed analysis of subunit–subunit interactions (Onrust
et al., 1995, and references therein) (Figure 1). As mentioned above, it has been proposed that the dual core
polymerases in pol III holoenzyme mediate the coordinated synthesis of the leading and lagging strands at
the replication fork (Johanson and McHenry, 1984; Wu
Cell
6
Table 1. Subunit composition of DNA Polymerase III holoenzyme
Subunit
Molecular Mass
(kDa)
Function
a
e
u
129.9
27.5
8.6
DNA polymerase
39 to 59 exonuclease
Stimulates e exonuclease
t
71.1
Dimerizes core
Binds g complex
g
d
d9
x
c
47.5
38.7
36.9
16.6
15.2
Binds
Binds
Binds
Binds
Binds
b
40.6
Sliding clamp
ATP
to b
to g and d
to SSB
to x and g
Subassemblies

 Core




 Pol III9









 Pol III*


 g Complex

 (DNA-Dependent ATPase) 




Adapted from Kelman and O’Donnell (1995).
et al., 1992). The single g complex in pol III holoenzyme
presumably serves to load b clamps onto both strands.
It is likely that the loading of a single b clamp is sufficient
for processive leading strand synthesis. However, as
described in greater detail below, the cycling of core
polymerase during lagging strand synthesis necessitates the loading of a clamp for each Okazaki fragment.
Other stable subassemblies of pol III holoenzyme have
also been purified from E. coli. These include the pol III
core, the g complex, and pol III9 (pol III* without the g
complex). It is not known whether these subassemblies
represent artifacts of purification or whether they have
specific roles in replication apart from those played by
the holoenzyme. It is also possible that they function in
other intracellular processes such as repair or recombination.
One of the major advances in understanding holoenzyme function came from studies of the b processivity
factor. When pol III is associated with the b clamp, its
processivity increases from about 10 nucleotides polymerized per binding event to over 50,000 nucleotides
polymerized per binding event (Mok and Marians, 1987).
The overall rate of polymerization increases from z20
to z750 nucleotides per second (Kelman and O’Donnell,
1995). Thus, the presence of the b clamp is absolutely
Figure 1. The Pol III* Subassembly of DNA Pol III Holoenzyme
Adapted from Onrust et al. (1995).
essential for the efficient duplication of the large E. coli
chromosome. In an elegant series of biochemical experiments, it was established that the b clamp is associated
with the DNA via a unique topological linkage. When b
dimers were assembled onto singly nicked circular DNA,
the resulting DNA–protein complexes were observed to
be extremely stable, dissociating with a half time of 72
min under physiological conditions (Stukenberg et al.,
1991; Yao et al., 1996). However, linearization of the
DNA with a restriction enzyme resulted in the rapid dissociation of b from the DNA, suggesting that b dimers
are capable of sliding freely on the DNA and can slip
off the ends of linear molecules. This general picture
gained support by the observation that the stability of
b dimers on linear DNA could be increased by the presence of sequence-specific DNA-binding proteins that
blocked the path to the DNA ends (Stukenberg et al.,
1991). These and other experiments led to the prediction
that the b dimer encircles the DNA. Subsequent X-ray
diffraction studies showed that the b dimer is indeed
ring-shaped and possesses a central cavity large
enough to accommodate a DNA duplex (Kong et al.,
1992). Based upon biochemical studies, the processivity
factors of the bacteriophage T4 and eukaryotic replication machines (gp45 and proliferating cell nuclear antigen [PCNA], respectively) were also predicted to be
ring-shaped structures capable of sliding along the DNA
(Alberts, 1987; Herendeen et al., 1992; Tinker et al., 1994;
Yao et al., 1996). In the case of PCNA, this prediction
has recently been confirmed by X-ray crystallographic
studies (Krishna et al., 1994).
Figure 2 outlines the current picture of the elemental
steps involved in the synthesis of an Okazaki fragment
by the pol III replication machine. The first step in the
sequence is the loading of the b clamp at a primer
terminus. The g complex, which functions as the clamp
loader, has been reconstituted from purified components, and some aspects of its mechanism of action are
beginning to emerge. It has been established that the
d subunit of the clamp loader is responsible for binding
the b dimer during the loading process (Naktinis et al.,
1995). Interestingly, the isolated d subunit can bind b in
the absence of ATP, while binding of the complete g
complex to b is almost completely ATP dependent. This
Minireview
7
Figure 2. Synthesis of Okazaki Fragments by Pol III Holoenzyme
during Lagging Strand Replication
For simplicity, the diagram shows only pol III core and the g complex.
Other components of the pol III holoenzyme, including the t subunit
and the second pol III core molecule that mediates leading strand
synthesis, are omitted. The timing of ATP hydrolysis is speculative.
has led to the hypothesis that the d subunit is normally
buried, but becomes exposed for interaction with b as
a result of an ATP-induced conformational change in
the g complex. The predicted conformational change
has been detected by analyzing changes in the sensitivity of the g complex to proteases upon binding ATP.
The clamp loader specifically recognizes primer termini
and transfers the bound b dimer onto the DNA in a
reaction that requires ATP hydrolysis. It has been demonstrated that the clamp loader is a DNA-dependent
ATPase whose activity is maximal in the presence of
both b and a primer terminus (Kelman and O’Donnell,
1995). One reasonable model is that hydrolysis of ATP
induced by DNA binding causes the d subunit to retract
again, releasing the b dimer onto the DNA. ATP hydrolysis might also reduce the affinity of the clamp loader
for the DNA facilitating its dissociation from the primer
terminus.
How the ring shaped b dimer is slipped onto the DNA
is an interesting problem that has not yet been solved.
One possibility (depicted in Figure 2) is that the binding
of the b dimer to the clamp loader breaks one (or both)
sets of contacts that hold the two b subunits together,
thus opening the protein ring. In this scenario, the ring
would close again when b is released from the loader
at the primer terminus. Alternatively, transient opening
of the ring may be coupled directly to the hydrolysis of
ATP. Given the molecular tools now available, the answer to this interesting mechanistic puzzle may soon
be forthcoming.
The second step in Okazaki strand synthesis is the
association of pol III core in the holoenzyme with the b
clamp to form a processive polymerase. In the absence
of DNA, the core polymerase appears to have a relatively
low affinity for b dimers. However, once the clamp is
placed at the primer terminus, the stability of the complex between core polymerase and the clamp is dramatically increased (Naktinis et al., 1996, and references
therein). The basis for this enhanced stability of the
core–b complex at a primer terminus is not fully understood, but could be due to extra contacts between polymerase and DNA or to a change in the structure of
the polymerase that augments the favorable contacts
between polymerase and clamp. Whatever its physical
basis, it is the stability of this complex that explains the
ability of pol III to polymerize thousands of nucleotides
without dissociating from the template.
Although tightly bound to the b clamp during processive DNA synthesis, the core polymerase suddenly
loses its affinity for the b clamp when it reaches the
end of the template and encounters the 59 end of the
previously synthesized Okazaki fragment. The polymerase then dissociates from the DNA leaving the b clamp
behind (Stukenberg et al., 1994). This third step in the
reaction sequence is clearly central to the process of
Okazaki strand synthesis, since it allows the polymerase
to cycle to the next primer. It is not yet clear what signal
is recognized by the polymerase to cause it to switch
out of the processive protein configuration upon completion of an Okazaki fragment. However, in vitro studies
indicate that this property is intrinsic to the pol III holoenzyme and does not require any accessory factors (Stukenberg et al., 1994). The T4 DNA polymerase holoenzyme is similarly programmed for rapid disassembly
upon completion of Okazaki strand synthesis (Hacker
and Alberts, 1994; reviewed by Stillman, 1994). Presumably, some structural feature at a nick (e.g., the abutting
59 terminus) induces a structural transition in the polymerase that breaks the contacts with the sliding clamp
and facilitates dissociation from the DNA.
What happens to the b clamps that are left behind on
the DNA when pol III dissociates from the competed
nascent strand? Given that the number of b dimers per
cell is an order of magnitude less than the number of
Okazaki fragments produced during the replication of
the E. coli genome (Kornberg and Baker, 1992), there
must be a mechanism to reuse such abandoned clamps.
In vitro studies suggest that the g complex itself is capable of catalyzing the efficient dissociation of b dimers
from the DNA (i.e., the clamp loader is also a clamp
unloader) (Stukenberg et al., 1994; Naktinis et al., 1996).
Whether the g complex functions as a clamp loader or
unloader is probably modulated by the DNA structure
to which it is bound. When the g complex is bound
tightly to a primer terminus, it adopts a conformation
capable of efficiently coupling ATP hydrolysis to the
transfer of b clamps onto the DNA. When bound to other
DNA structures or free in solution, the g complex may
adopt a different conformation in which this coupling is
lost. The latter conformation may still be capable of
catalyzing the opening and closing of the protein ring,
thus allowing the rapid equilibration of b clamps on and
off the DNA.
Cell
8
The foregoing discussion suggests a possible paradox. If the g complex can function as a clamp unloader,
what prevents it from removing b clamps associated
with pol III core during the processive stage of DNA
synthesis? If such unloading events were to occur at a
significant frequency, the overall efficiency of lagging
strand synthesis would be greatly reduced. Work by
Naktinis et al. (1996) suggests that E. coli has evolved
an economical solution to this problem. Using both biochemical and genetic approaches, Naktinis et al. demonstrated that both the pol III core and the g complex
bind to the same face of the b ring via contacts near
the C-termini of each b monomer. As a consequence of
this overlap of binding sites, pol III core and the g complex cannot be bound to the b clamp at the same time.
During processive DNA synthesis, the stable association
of b with pol III core prevents access by the g complex
and thus effectively prevents premature unloading of
the clamp. It is only after completion of an Okazaki
fragment, when the pol III core dissociates from b, that
the g complex can access the clamp and mediate unloading.
It is apparent from the foregoing discussion that the
smooth functioning of the pol III replication machine
depends upon a number of specific protein–protein and
protein–DNA interactions. These interactions ensure
that the events required for synthesis of each Okazaki
fragment take place in the proper order and are completed rapidly. One reason for the speed of the machine
is that all of the reacting components are held in close
proximity by protein–protein interactions. Thus, even
though the lagging strand polymerase must constantly
dissociate from the termini of completed Okazaki fragments, the pol III holoenzyme is held at the fork by the
leading strand polymerase, which remains tethered to
the DNA via a b clamp throughout chromosomal DNA
replication. The physical proximity of the polymerase
active site to a newly synthesized RNA primer facilitates
the cycling of the polymerase. Similarly, the presence
of a g complex within the holoenzyme ensures the rapid
assembly of b clamps on newly synthesized primers, as
well as their rapid disassembly from completed DNA
strands. Although it has not yet been possible to measure the time required for the intramolecular cycling of
pol III core from completed strand to nascent primer, it
has been estimated from in vitro studies of intermolecular transfer that the cycle time is considerably less than
1 s (Stukenberg et al., 1994). Given that it requires 1–2 s
to complete the synthesis of an Okazaki fragment, it is
clear that the performance of the machine is not limited
by the time required for polymerase cycling.
Much of what has been learned of the mechanism
of pol III action may be generally applicable to other
replicases and even to processes unrelated to DNA replication. Both T4 and E. coli utilize dimeric polymerases
for coordinated synthesis of leading and lagging strand
DNA (Alberts, 1987; Johanson and McHenry, 1984). Further characterization of the eukaryotic replication mechanisms may also reveal physical coupling of the leading
and lagging strand polymerases. As already noted, both
the T4 and eukaryotic replication machines make use
of ring-shaped homotrimeric sliding clamps (gp45 and
PCNA) similar in overall structure to the b dimer, and
both have clamp loaders that function like the g complex
(Alberts, 1987; Kornberg and Baker, 1992; Stillman,
1994). Interestingly, work in the past several years has
uncovered cases in which components of the replicase,
particularly the sliding clamps, interact with proteins not
involved in DNA synthesis. In the case of bacteriophage
T4, for example, it has been demonstrated that the gp45
sliding clamps abandoned during DNA replication can
serve as mobile enhancer proteins to activate the RNA
polymerase responsible for transcribing the late genes
of the virus (Herendeen et al., 1992). This mechanism
explains how the switch from early to late gene expression during T4 infection is coupled to the onset of DNA
replication. In eukaryotes it has been shown that a significant fraction of PCNA in the cell is present in complexes
with cyclin-dependent kinases and the p21 kinase inhibitor, suggesting that the protein may play a role in linking
DNA replication to other processes in the cell cycle
(Xiong et al., 1992). Thus, it appears that cells have found
additional uses for the complex machinery that originally
evolved to duplicate the genome. It would not be surprising if other examples are uncovered in the course of
future studies of pol III function.
Selected Reading
Alberts, B.M. (1987). Phil. Trans. R. Soc. (Lond.) B 317, 395–420.
Hacker, K.J., and Alberts, B.M. (1994). J. Biol. Chem. 269, 24221–
24228.
Herendeen, D.R., Kassavetis, G.A., and Geiduschek, E.P. (1992).
Science 256, 1298–1303.
Johanson, K.O., and McHenry, C.S. (1984). J. Biol. Chem. 259, 4589–
4595.
Kelman, Z., and O’Donnell, M. (1995). Annu. Rev. Biochem. 64,
171–200.
Kong, X.-P., Onrust, R., O’Donnell, M., and Kuriyan, J. (1992). Cell
69, 425–437.
Kornberg, A., and Baker, T. (1992). DNA Replication, Second Edition
(New York: W.H. Freeman and Company).
Krishna, T.S.R., Kong, X.-P., Gary, S., Burgers, P., and Kuriyan, J.
(1994). Cell 79, 1233–1244.
Mok, M., and Marians, K.J. (1987). J. Biol. Chem. 262, 16644–16654.
Naktinis, V., Onrust, R., Fang, L., and O’Donnell, M. (1995). J. Biol.
Chem. 270, 13358–13365.
Naktinis, V., Turner, J., and O’Donnell, M. (1996). Cell 84, this issue.
Onrust, R., Finkelstein, J., Turner, J., Naktinis, V., and O’Donnell, M.
(1995). J. Biol. Chem. 270, 13366–13377.
Stillman, B. (1994). Cell 78, 725–728.
Stukenberg, P.T., Studwell-Vaughan, P.S., and O’Donnell, M. (1991).
J. Biol. Chem. 266, 11328–11334.
Stukenberg, P.T., Turner, J., and O’Donnell, M. (1994). Cell 78,
877–887.
Tinker, R.L., Kassavetis, G.A., and Geiduschek, E.P. (1994). EMBO
J. 13, 5330–5337.
Wu, C.A., Zechner, E.L., Hughes, A.J., Franden, M.A., McHenry, C.S.,
and Marians, K.J. (1992). J. Biol. Chem. 267, 4064–4073.
Xiong, Y., Zhang, H., and Beach, D. (1992). Cell 71, 505–514.
Yao, N., Turner, J., Kelman, Z., Stukenberg, P.T., Dean, F., Shechter,
D., Pan, Z.-Q., Hurwitz, J., and O’Donnell, M. (1996). Genes to Cell,
in press.